SUMMARY An obstacle to studying human cancer is the limited availability of models with human stroma and immune cells. This is particularly relevant for head and neck squamous cell carcinomas (HNSCC) given the pivotal role the immune system plays in their development. An unmet need is an investigation of how cancer stem cells (CSCs) modulate immunity, because it is critical to understand how CSCs harness oncogenic signaling to evade immune surveillance. To address these outstanding questions, we will use two unique tools: 1) CSCs derived from a panel of patient HNSCCs propagated in mice, and 2) a humanized mouse (HM) model that enables studying tumor-stromal interactions. We have defined CSCs as ALDH+CD44high in multiple patient cases, and documented that SOX2 is responsible for key CSC features such as growth, invasion, and resistance to therapy (1). Further, we observed an association between SOX2 and tumor expression of the ligand PD-L1, which activates PD-1 in T cells leading to immune escape. Additionally, SOX2 modulates SOX9, which regulates the other PD-1 ligand, PD-L2. After treatment of HNSCC HM mice with a PD-1 inhibitor, we documented resistance associated with an increase in PD-L2; thus, we propose that alternate ligands are used by HNSCC as a resistance mechanism. We initially developed a mismatched HM (mHM; tumor and hematopoietic stem cells [HSCs] from different sources), and now have generated an autologous HM (aHM; tumor, HSCs and mesenchymal stem cells [MSCs] from the same patient). Tumors grown in HM more closely resemble the originator tumor than those grown in non-HM mice, and the drift in gene expression caused by prior passaging was partially reversed. Signaling in key immune and stroma pathways was more prominent and closely resembled the originator patient in HM vs. non-HM models. The immune cells mounted effective tumor-specific immune responses, mediated by human immune cells including T cells. Two salient and under- studied issues limit the wider application of HM: tumor-immune mismatch and T cell education. Tumor and immune matching can affect the faithfulness of immune response, and that can only be appreciated by comparing mHM and aHM. HM are subject to xenogeneic education of human T cells by mouse thymic cells, resulting from the lack of a functional human thymus in such models. The goals of this proposal are to: 1) study the regulation by SOX2 and SOX9 of PD-1 ligands; 2) understand how SOX2 and SOX9 affect CSC properties; 3) generate and characterize aHM and mHM from 10 HNSCC patients; 4) test if SOX2 and SOX9 are involved in resistance to PD-1 inhibitors by testing them on HM; and finally, we will 5) generate thymic epithelium from the same patient's HSCs that will result in HM with a fully autologous HNSCC, thymus and immune system, thus enabling immune cell education in a strictly human context. This project will lead to a deeper and more mechanistic understanding of the interplay between HNSCC CSCs and the immune/stroma systems, and ultimately, improve precision medicine.